Jaivardhan Kapoor
Hello! I am Jai.
I’m a third-year PhD candidate at the University of Tübingen and a scholar at the International Max Planck Research School for Intelligent Systems (IMPRS-IS). I am supervised by Prof. Dr. Jakob Macke and Prof. Dr. Christian Baumgartner. My research focuses on building diffusion-based generative models for applications in medical imaging and neuroscience.
I completed my B.Tech. and M.Tech. in Electrical Engineering from the Indian Institute of Technology Kanpur. In my master’s thesis supervised by Prof. Vipul Arora, I worked on a neural-network based generative model for the XY model using normalizing flows (pdf). The XY model is a statistical model for studying magnetic properties of matter in Physics. I have done research internships at the MPI for Intelligent Systems under Prof. Isabel Valera, and Aalto University under Prof. Harri Lahdesmaki.
PhD Projects.
One of my central works in my PhD work is building generative models for 3D brain MRIs.
- M3AE for unconditional generation, NeurIPS 2022 Medical Imaging Workshop. I use additive and diffeomorphic transformations with a VAE to generate 3D brains from a fixed template.
- MRExtrap for conditional aging, Medical Imaging and Deep Learning 2024. I use linearity of the autoencoder latent space to flexibly age brain MRIs from previous observations of a subject. (Long paper coming soon!)
I build generative models for neural timeseries data as well.
- Latent Diffusion for Neural Spiking Data, Neurips’24 Spotlight. We develop a latent diffusion model to generate realistic multi-channel neural spiking data conditioned on behavioural covariates.
I also develop diffusion model-based methods for uncertainty estimation and anomaly detection.
- Unsupervised Anomaly Detection using Aggregated Normative Diffusion ANDi, Preprint. We develop a simple, efficient and highly effective way to use unconditional diffusion models for unsupervised anomaly detection in medical imaging data. Key insight is that multiscale gaussian noise for diffusion works really well for this task.
Reach out to me!
I am looking for collaborators and company internship opportunities, specifically in diffusion generative models for various data domains and applications. Reach out to me on X/Twitter, Linkedin, or at jaivardhan.kapoor@uni-tuebingen.de.